Letters in Applied Microbiology 2000, 30, 468472

A comparison of the Bioscreen method and microscopy for the

determination of lag times of individual cells of Listeria
monocytogenes
Y. Wu1, M.W. Griffiths1 and R.C. McKellar2
1

Department of Food Science, University of Guelph and 2Food Research Program, Agriculture and Agri-Food Canada,
Guelph, Ontario, Canada
505: received 1 January 2000, revised 12 January 2000 and accepted 6 March 2000
Y . W U , M . W . G R I F F I T H S A N D R . C . M C K E L L A R . 2000. Lag phase durations (tLag) of individual
Listeria monocytogenes cells were analysed using the NightOwl Molecular Imaging System,
and results were compared with mean individual cell lag times (tL) obtained from the
detection time (td) method using Bioscreen. With Bioscreen, an average tL of 6
39 0
89 h
was obtained from ve separate experiments. With the NightOwl method, an average tLag of
2
73 0
06 h was obtained from three experiments consisting of eight total replicates. Lag
values from the NightOwl and Bioscreen are related by the equation: tLag tL DT,
where DT is the doubling time. The equivalent tLag mean value for the Bioscreen method
was 7
11 0
84 h. Individual lag times measured by both methods were normally
distributed (r2 for Bioscreen and NightOwl ranged from 0
951 to 0
999 and from 0
884 to
0
982, respectively). The results suggest that the NightOwl method can provide accurate
estimates of individual cell lag times, which will facilitate the development of combined
discrete continuous models for bacterial growth.

INTRODUCTION

Predictive microbiology, the use of mathematical models to

describe the growth and death of foodborne micro-organisms, has been an area of considerable activity over the last
decade. It is based upon the premise that the responses of
populations of micro-organisms to environmental factors
are reproducible, and that it is possible, from past observations, to predict the responses of micro-organisms by considering environments in terms of identiable dominating
constraints. Proponents claim that predictive microbiology
offers many benets to the practice of food microbiology,
and there is growing international interest in its use (Ross
and McMeekin 1994).
Kinetic models are perhaps the most useful, since they
can be used to predict changes in microbial numbers with
time, even if one (or more) of the controlling factors affecting growth is changing e.g. during a chilled distribution
chain (McClure et al. 1993). Many kinetic models have
been developed by tting growth curves of viable count
data obtained from cultures grown in liquid media
(Buchanan and Phillips 1990; Wijtzes et al. 1993). Viable
Correspondence to: M.W. Grifths, Department of Food Science, University of
Guelph, Guelph, Ont., Canada, N1G 2W1 (e-mail: mgrift@uoguelph.ca).

count is a traditional, sensitive method for estimating the

microbial growth curve, but it is time-consuming and
labour-intensive (McClure et al. 1993; Dalgaard et al.
1994). The amount of data required to generate reliable
models has led some researchers to use simple, and often
indirect, methods of data collection, such as turbidimetry
in laboratory media, rather than the viable count method.
Determining bacterial growth rates in broth systems using
turbidimetric methods provides a rapid and inexpensive
means of modelling; however, the use of indirect methods
for growth curve generation may result in generation times
different from those determined by using viable counts
(Baranyi et al. 1993; Dalgaard et al. 1994).
Turbidimetric methods such as the Bioscreen method
have been used to generate kinetic data for modelling by
tting non-linear regression functions to optical density
(OD) data (McClure et al. 1993; Stephens et al. 1997).
Determinations of the specic growth rate (m) and population lag phase duration (l) are often difcult with this
approach. Thus attempts have been made to use detection
times (td) (Cuppers and Smelt 1993). The td for a turbidimetric instrument can be dened as the time required for
an initial measurable increase in OD. When td values are
plotted against the corresponding inoculum size m can be
calculated as the negative reciprocal of the slope of the
= 2000 The Society for Applied Microbiology

LISTERIA MONOCYTOGENES LAG TIMES

regression line (Cuppers and Smelt 1993). The mean individual cell lag times (tL) can be calculated subsequently as
the difference between the predicted td based on the m, and
the observed td. This approach has been used recently to
determine tL for L. monocytogenes (McKellar 1998;
McKellar and Knight 2000).
With the combination of microscopy and imaging, direct
observation of single cell growth has become possible, providing a direct observation of the growth kinetics of individual bacterial cells. Lag phase can be obtained by
determining when the rst doubling occurs, and growth
rate can be estimated by counting the doubling time of the
cells. Few studies have attempted to derive kinetic models
using microscopy techniques. The behaviour of individual
cells in foods is poorly understood. In spoilage situations
this may not be important; however, some food-borne
pathogens can cause illness from only a few cells (Mackey
and Gibson 1997). Thus, determining the lag time of single
cells would provide valuable information for risk analysis.
Therefore, the purpose of this study was to determine the
lag phase of individual cells using microscopy, and to compare the distribution of individual cell lag times with that
obtained using the Bioscreen method.
MATERIALS AND METHODS
Strains and culture conditions

Listeria monocytogenes Scott A (human clinical isolate) was

obtained from the culture collection of the Food Research
Program (Guelph, Ont. Canada). API Listeria spp. identication strip (BioMerieux Canada Inc., St Laurent, PQ,
Canada) was used to conrm the identity of the culture.
The culture was grown for 24 h at 30  C in tryptic soy
broth (TSB; Difco Laboratories, Detroit, MI, USA). Stock
cultures were prepared in TSB plus 15% glycerol (BDH
Inc., Toronto, Ont., Canada) and 0
3 ml subsamples were
frozen in cryovials at 25  C.
The contents of one cryovial were transferred to 10 ml of
TSB, and incubated for 24 h at 30  C in a shaking water
bath (Model 3100, New Brunswick Scientic, Edison, NJ,
USA) at 1500 rev min1. The culture (0
1 ml) was transferred to 10 ml fresh TSB and incubated at 30  C for 24 h
when the cell density was approximately 108 cfu ml1. For
the Bioscreen, the resulting culture was used as the inoculum for experiments. For microscopy, the resulting culture
was diluted 1 : 10 in TSB, and the diluted suspension
(approximately 107 cfu ml1) was used to prepare the slide.
Bioscreen technique
Growth experiments. Serial twofold dilutions of the inoculum were made using fresh TSB to obtain a range of dilu-

469

tions representing approximately 0105 cfu ml1. From

each of the twofold dilutions, 350 ml was transferred to
wells (40 wells per dilution) of a Bioscreen plate
(Labsystems, Helsinki, Finland). The lled plates were
placed in the Bioscreen for analysis. Measurements were
taken using a wide band lter, with preshaking at medium
intensity for 10 s prior to OD reading, at an incubation
temperature of 30  C. Measurements were taken every 4
min for 25 h. Results were reported as td (h) and dened as
the time required for the Bioscreen to record a 0
05
increase in OD from the initial value.
Viable cells were enumerated for each twofold TSB dilution by spread plating appropriate serial dilutions (0
1 ml)
in duplicate onto tryptic soy agar plates (TSA, Difco
Laboratories). The plates were incubated at 30  C for 48 h
and counts were determined using a Quebec Counter
(American Optical Co., Model 15, Buffalo, NY, USA).
Modelling. Calculation of tL (mean individual cell lag
times) was performed as described previously (McKellar
1998; McKellar and Knight 2000). Briey, plots of td
(detection time) obtained from serial dilutions of the original inoculum against ln cfu ml1 were used to calculate m
from the slope using the equation:

m 1=slope

The calculated m was then used in the heterogeneous

population model (McKellar 1997) to predict the time
required to detect growth from a dened number of cells.
It was assumed that the dilution giving the largest td was
equal to 1 cfu well1 (or ln cfu well1 0). Simulated
values for td underestimated the actual td by an amount
equivalent to tL. Replicate values of tL and the standard
deviation (S.D.L) were calculated from ve independent
40-well trials by subtracting the simulated value for td from
the experimental td values.
For each of the Bioscreen experiments, frequency distributions of individual tL values from replicate wells were
calculated using the non-linear regression function of
Prism Version 3
0 (GraphPad Software for Intuitive
Science, San Diego, CA, USA). Bin widths were set at 0
7
h intervals, and the number of wells corresponding to each
bin was determined. A normal distribution was then tted
to the resulting frequency distribution using Prism.
Microscopic technique
Slide preparation. Glass double cavity slides (75  25 mm,
18 mm diameter cavities, 1
75 mm thickness) and coverslips
(22 cm2) (VWR Canlab, Mississauga, Ont., Canada) were
autoclaved before use.

= 2000 The Society for Applied Microbiology, Letters in Applied Microbiology, 30, 468472

470 Y . W U E T A L .

Two hundred and fty microlitres of molten TSA were

pipetted into each cavity of the slide. The agar in each cavity was covered with a coverslip, and pressure was applied
to obtain a at, smooth surface. The slide was left for 5
min to allow the agar to solidify. The coverslip was
removed, 2 ml of sample suspension was pipetted onto the
surface of the agar, and another coverslip was applied. The
slide was placed in a sealed Petri dish and incubated at 30

C.
Image capture and analysis. Images were acquired and
evaluated using the NightOwl LB 981 Molecular Imaging
System (EG & G Berthold, Bad Wildbad, Germany).
Microscopic images were obtained using the CCD camera
of the NightOwl system mounted on a microscope
(Olympus BH2, Capsen Ltd, Markham, Ont., Canada).
The growth of the target cells in two or three different
elds of view for each slide was visualized using phase-contrast microscopy. Images of each area were captured at
time intervals of 0, 1, 1
5, 2, 2
25, 2
5, 2
75, 3, 3
25 and
3
5 h. After each observation, the slide was replaced in a
sealed Petri dish and incubated at 30  C. The number of
newly divided cells was counted and these cells were not
counted again in later captured images.
Three experiments were conducted; in experiment 1,
two replicate areas were viewed per slide; in experiments 2
and 3, three replicate areas were viewed.

Modelling. For each microscopy experiment, the number of

cells that had newly divided during each 0
25 h time interval was determined and used as the frequency distribution.
A normal distribution was then tted to the resulting frequency distribution using Prism.

Fig. 1 Examples of normal distribution curves from microscopy.

(*) replicate 2 of experiment 2 (r2 0
984) and the (W)

Bioscreen experiment 1 (r2 0
997)

RESULTS

The average tL from the Bioscreen was 6
39 0
89 h

(Table 1). All the data t a normal distribution well; r2
values ranged from 0
951 to 0
999 (Fig. 1). Individual
values for S.D. from each experiment ranged from 0
567 to
0
991.
With the microscopic method, eight replicates were
obtained from three experiments. The total cell number in
each eld of view was highly variable (from 24 to 166 cells)
among replicates because cells were not evenly distributed
on the agar (Table 2). The average lag phase duration
observed microscopically (tLag) was 2
73 0
06 h. Non-linear regression analysis indicated that all the replicates were
normally distributed; r2 values ranged from 0
884 to 0
984

Table 1 Results from Bioscreen

Experiment number

tL

S.D.

R2
DT
tLag

Average

3
74
0
567
0
997
0
677
5
57

6
21
0
637
0
999
0
672
6
86

7
19
0
646
0
993
0
680
7
82

6
81
0
991
0
964
0
680
7
48

6
99
0
793
0
951
0
680
7
80

6
39 0
887
0
727

7
11 0
843

tL mean individual lag times from Bioscreen data (h).

S.D. standard deviation.

DT doubling time (h).

tLag tL DT.

= 2000 The Society for Applied Microbiology, Letters in Applied Microbiology, 30, 468472

LISTERIA MONOCYTOGENES LAG TIMES

(Fig. 1). From the images it was observed that not all cells
divided within the time period of the experiment although
the number of such cells was less than 5% of the total cell
population.
The tL derived using the Bioscreen method assumes that
individual cells start growing at maximal growth rate
immediately after adaptation (McKellar 1998; McKellar
and Knight 2000). The time at which the rst cell was
observed microscopically to be dividing (tLag) includes the
adaptation period (tL) and the time required for the cell to
double (DT). Thus tLag is related to tL by equation 2.
tLag tL DT

DT was calculated from m using equation 3:

DT ln2=m

Calculations for tLag from the Bioscreen data are given in

Table 1.

DISCUSSION

The results of the present study show that mean individual

cell lag times obtained either by the Bioscreen method or
by microscopy follow normal distributions (Fig. 1). There
have been few studies published on the distribution of bacterial cell lag times. Baranyi (1998) and Baranyi and Pin
(1999) proposed that lag times were exponentially distributed. However, other important physiological parameters,
such as growth rate, were found to be normally distributed
(Kelley and Rahn 1932; Kubitschek 1966; Rubinow 1980).
In previous studies employing the Bioscreen, it was
assumed that tL values were normally distributed, and, by
using this assumption, the decreased variability in tL
between replicate wells with >1 cell well1 initial count
was correctly predicted (McKellar 1998; McKellar and

471

Knight 2000). These same studies showed that the assumption of exponential distributions for tL led to predictions
which did not agree with experimental ndings (McKellar
1998; McKellar and Knight 2000). Further work is
required to more completely classify the distribution of cell
physiological properties.
The lag times determined by the Bioscreen method in
this study were slightly greater than were observed previously (McKellar 1998; McKellar and Knight 2000); however, the S.D. values are comparable. The tL values
obtained by the Bioscreen method were also longer than
those found microscopically. The reason for this difference
is unknown. However, Baranyi et al. (1993) and Dalgaard
et al. (1994) have suggested that the use of indirect methods for growth curve generation might result in generation
times different from those determined using viable counts.
There may also be differences between growth of microorganisms in solid media on slides and in liquid media due
to O2 availability and other factors. Further research is
needed to resolve this issue.
Using microscopy to determine the lag, the zero time
point was counted when the rst image was captured and
there was a slight delay between this rst capturing time
and the actual inoculation time. This would result in the
actual tLag being slightly longer than the calculated tLag.
Since this slight time difference was about 5 min in each
replicate and was constant for each experiment, it was
ignored as a system error.
From the microscopic images, it was observed that some
cells did not divide within the time period of the experiment. These cells can be considered dead, metabolically
inactive or injured. This is one advantage of the
microscopic method since dead, injured or living cells can
be distinguished based on their ability to divide. Thus,
cells under stress can be directly observed and their recovery can be studied in vivo. Furthermore, data on the
kinetics of growth at other points in the growth cycle can
be obtained.

Table 2 Results from microscopy

Experiment 1

tLag

S.D.

R2
Total cells*

Experiment 2

Experiment 3

Replicate 1

Replicate 2

Replicate 1

Replicate 2

Replicate 3

Replicate 1

Replicate 2

Replicate 3

Average

2
80
0
383
0
884
58

2
84
0
306
0
932
45

2
71
0
366
0
961
24

2
79
0
499
0
963
37

2
65
0
414
0
984
74

2
76
0
429
0
982
166

2
63
0
408
0
962
81

2
67
0
361
0
977
71

2
73 0
06
0
396

69
5 43
6

*Number of total divided cells counted in each area tLag, mean individual lag times from microscopy data (h).
= 2000 The Society for Applied Microbiology, Letters in Applied Microbiology, 30, 468472

472 Y . W U E T A L .

ACKNOWLEDGEMENTS

The authors would like to thank Agriculture and AgriFood Canada, through their Science Horizons program,
the Dairy Farmers of Ontario, and the Natural Science and
Engineering Research Council of Canada for nancial support of this project.
REFERENCES
Baranyi, J. (1998) Comparison of stochastic and deterministic concepts of bacterial lag. Journal of Theoretical Biology 192, 403
408.
Baranyi, J. and Pin, C. (1999) Estimating bacterial growth parameters by means of detection times. Applied and Environmental
Microbiology 65, 732736.
Baranyi, J., Roberts, T.A. and McClure, P. (1993) A non-autonomous differential equation to model bacterial growth. Food
Microbiology 10, 4359.
Buchanan, R.L. and Phillips, J.G. (1990) Response surface model
for predicting the effects of temperature, pH, sodium chloride
content, sodium nitrite concentration and atmosphere on the
growth of Listeria monocytogenes. Journal of Food Protection 53,
370376, 381.
Cuppers, H.G.A.M. and Smelt, J.P.P.M. (1993) Time to turbidity measurement as a tool for modeling spoilage by
Lactobacillus. Journal of Industrial Microbiology 12, 168171.
Dalgaard, P., Ross, T., Kamperman, L., Neumeyer, K. and
McMeekin, T.A. (1994) Estimation of bacterial growth rates
from turbidimetric and viable count data. International Journal
of Food Microbiology 23, 391404.
Kelley, C.D. and Rahn, O. (1932) The growth rate of individual
bacterial cells. Journal of Bacteriology 23, 147153.

Kubitschek, H.E. (1966) Normal distribution of cell generation

rates. Nature 209, 10391040.
Mackey, B.M. and Gibson, G.R. (1997) Escherichia coli: From
farm to fork and boyond. Society for General Microbiology
Quarterly 24, 5557.
McClure, P.J., Cole, M.B., Davies, K.W. and Anderson, W.A.
(1993) The use of automated tubidimetric data for the construction of kinetic models. Journal of Industrial Microbiology 12,
277285.
McKellar, R.C. (1997) A heterogeneous population model for the
analysis of bacterial growth kinetics. International Journal of
Food Microbiology 36, 179186.
McKellar, R.C. (1998) A Discrete Adaptation Model Describing the
Lag Phase of Listeria Monocytogenes. Eighth International
Symposium on Microbial Ecology. Halifax, NS: McCurdy
Printing.
McKellar, R.C. and Knight, K.P. (2000) A combined discretecontinuous model describing the lag phase of Listeria monocytogenes. International Journal of Food Microbiology 54, 171180.
Ross, T. and McMeekin, T.A. (1994) Predictive microbiology.
International Journal of Food Microbiology 23, 241264.
Rubinow, S.I. (1980) Cell kinetics. In Mathematical Models in
Molecular and Cell Biology ed. Segel, L.A. pp. 502522.
Cambridge: Cambridge University Press.
Stephens, P.J., Joynson, J.A., Davies, K.W., Holbrook, R.,
Lappin-Scott, H.M. and Humphrey, T.J. (1997) The use of an
automated growth analyser to measure recovery times of single
heat-injured Salmonella cells. Journal of Applied Microbiology
83, 445455.
Wijtzes, T., McClure, P.J., Zwietering, M.H. and Roberts, T.A.
(1993) Modelling bacterial growth of Listeria monocytogenes as a
function of water activity, pH and temperature. International
Journal of Food Microbiology 18, 139149.

= 2000 The Society for Applied Microbiology, Letters in Applied Microbiology, 30, 468472